spective study on the island of Kauai by Yerushalmy and his colleagues (31). Their methods were summarized earlier, but one detail is of special interest here. Yerushalmy, in an earlier report (61), had described the effects of prematurity on the association between birth interval and mortality. He gave Eastman (62) credit for pointing out that if a full-term infant is to be born within a year after the termination of a previous pregnancy, conception would have to occur within 3 months of that termination, whereas aborted fetuses or prematurely born infants, conceived 4 months or more after the previous delivery, could still be born within a year of the previous delivery. The high rates of mortality associated with prematurity would therefore artificially inflate the mortality rates of the 1-year-or-less birth interval group.
For this reason, they based their analysis of the Kauai data on pregnancy interval rather than birth interval. In addition, they calculated the relative mortality rates for the various fetal and child age groups in each pregnancy interval in order to bring out the differences more clearly. Their findings are shown in Figure 11, where it is clear that for all child age groups, death rates are highest in the shortest interval groups. For fetal and neonatal deaths, when biological factors are most important, rates are highest in the shortest interval group. They decrease to a minimum with an interval of around 2 years, then increase as the interval increases further, though never equaling the rates in the shortest interval. For postneonatal and early childhood mortality, when environmental factors would be most important, the association is consistent and almost linear: as intervals increase, chances of survival increase.
Yerushalmy had already examined the effect of birth interval on stillbirths (61). He studied all the births and stillbirths that occurred in the United States during the 5-year period 1937-1941. These were classified according to age of mother and parity, and the sample included 7,151,631 births, of which 211,079 were stillbirths. He analyzed the data on the basis of the assumption that, in general, for women in the same age group, the interval between births decreases with increasing parity, and thus provides, as he termed it, an indirect method of studying birth interval. While all of his data need not be reviewed in detail, it is pertinent here to mention two of his observations. First, he noted that when the stillbirth rates were tabulated according to age and parity, the minimum ("best") rates appear in increasingly older age groups as parity increases—as the probable interval between births increased, the stillbirth rates declined. Second, he observed
It is remarkable that although the level of the stillbirth rate is higher for nonwhite than for total births, and although the effect of the birth-order factor and the effect of the age-of-mother factor by themselves arc less pronounced among nonwhite than among total births, the percentage increase or decrease in the stillbirth rate which may be attributed to the